在Opal中构建动态智能体工作流。
内容来源：https://blog.google/innovation-and-ai/models-and-research/google-labs/opal-agent/
内容总结：
Opal平台近日完成重要升级，将传统静态工作流全面升级为智能体驱动模式。用户现可在"生成"步骤中选择智能体，系统将根据任务目标自主规划执行路径，动态调用网络搜索、视频生成等工具模型，显著降低复杂任务的操作门槛。
以故事创作场景为例，新版"视觉叙事"智能体可自主推演情节脉络、补充关键细节，实现从固定模板到动态叙事的跨越。在室内设计应用中，用户上传空间照片并描述风格倾向后，智能体不仅能生成初步设计方案，还能通过多轮对话精准捕捉用户偏好，甚至主动调研小众设计流派，使输出成果摆脱千篇一律的模板化呈现。
本次升级同步推出三大核心功能：
- 记忆存储：可跨会话记忆用户偏好，如品牌调性、风格倾向等信息
- 动态路由：支持基于条件判断的智能路径选择，如在商务简报场景中自动区分新老客户并采取相应调研策略
- 交互对话：智能体可主动发起追问或提供选项，确保信息完整性
平台在提升自动化水平的同时，仍保留传统工作流步骤以满足高精度原型设计需求。这种"智能体自主决策+人工实时调控"的双重模式，既降低了新用户上手难度，又为专业开发者提供了灵活构建空间。目前该升级已全面上线，预计将推动人机协作模式进入新阶段。
中文翻译：
在Opal中构建动态智能体工作流
今天，我们将Opal工作流从静态模型调用升级为智能体驱动。现在您无需手动选择模型，只需在“生成”步骤中选择智能体即可。该智能体步骤会根据您的目标主动规划最佳路径，自动调用合适的工具与模型（例如用于研究的网络搜索或用于视频生成的Veo），从而以更少的手动配置完成复杂任务。
以往创建故事书Opal时，需要预先设定页数和用户问题。如今，我们可以构建一个“视觉叙事者”Opal——其中的智能体步骤能自主判断所需细节，并主动建议情节走向，帮助您引导故事发展。这标志着从固定模板到实时创意决策驱动的动态独特叙事的转变。
从静态体验到交互式体验
以室内设计Opal为例。在引入智能体步骤前，室内设计Opal更像单向流程：上传图片、输入风格偏好、获取重新设计后的空间图像。而全新的智能体步骤让升级后的“房间风格设计师”Opal变得互动性十足，仿佛一位与您协作的设计伙伴。
上传一张空客厅的照片，描述您的中世纪现代风格构想。智能体将生成包含时代特色装饰与配色方案的初始概念。若效果未达预期，您可针对具体元素提供反馈。通过多轮对话迭代，智能体会逐步深化对您审美偏好的理解，甚至能研究小众设计子风格，最终生成独具个性而非千篇一律样板间的改造图像。
Opal之所以能实现此类交互体验，是因为智能体能理解您的目标、规划最优执行方案、在需要时主动征询意见，并调度最合适的模型与工具完成任务。
赋能Opal智能体的新工具
- 记忆功能：无论是用户姓名、风格偏好还是动态购物清单，您的Opal现在可以跨会话记忆信息。使用越频繁，Opal就越智能且更具个性化。例如在“视频创意构思”Opal中，智能体步骤会将用户品牌标识与偏好存储至记忆库，让您无需重复输入即可即时生成定制化视频创意。
- 动态路由：通过自定义逻辑为智能体设定多重执行路径，全面掌控工作流。只需描述判断条件，智能体便会在条件满足时智能跳转至对应步骤。例如在“高管简报”Opal中，智能体会根据会面对象是新客户还是老客户定制简报内容——为新客户搜索网络背景资料，为老客户调取内部会议纪要提供相关上下文。
- 交互式对话：当AI智能体需要补充信息时，智能体步骤可主动发起聊天以收集缺失内容，或在进入下一阶段前提供选项。以“房间风格设计师”Opal为例，若用户初始信息不足，Opal将持续提问或展示示例以供参考。
更多Opal构建方式
我们相信这种设计实现了双重优势：既能发挥AI智能体为实现目标而工作的强大能力，又保留了可随时自定义优化的分步工作流控制权。
在显著提升功能的同时，我们始终保持着Opal的简洁性。新用户将感受到Opal“开箱即用”的便捷——因为生成步骤中的智能体具备自我修正、记忆与优化的能力。而对于高阶用户和开发者，当需要高精度原型设计或严格逻辑时，标准固定步骤仍随时可用。通过弥合自动化与控制之间的鸿沟，我们正不断拓展创造力的边界。期待见证您基于智能体的全新Opal实践！
英文来源：
Build dynamic agentic workflows in Opal
Today we’re upgrading Opal workflows from static model calls to agentic intelligence. Instead of manually picking a model, you can now select an agent in the "generate" step. This agent step proactively determines the best path based on your goal, triggering the right tools and models (like Web Search for research or Veo for Video) to automate complex tasks with less manual configuration.
Previously, creating a storybook Opal required you to predefine page counts and user questions. Now, we can create a Visual Storyteller Opal where the agent step autonomously decides which details it needs and suggests plot points to help you direct where the story goes. This marks a shift from rigid formats to dynamic, unique narratives shaped by real-time creative decisions.
From static to interactive experiences
Let’s take an interior design Opal as an example. Before the agent step, an Interior Design Opal felt like a simple one-way process: upload a picture, input your style, and receive an image of your redesigned space. With the new agent step, your leveled up Room Styler Opal starts to feel interactive and more like another design partner you're collaborating with.
Upload a photo of your empty living room and describe your mid-century modern vision. The agent will generate an initial concept featuring era-specific decor and palettes. If it’s not quite right, you can provide feedback on specific elements. By iterating through this dialogue, the agent refines its grasp of your aesthetic and can even research niche design sub-styles to create redesigned images that feel uniquely yours, rather than a generic showroom template.
Opal can now create these interactive experiences because the agent understands your goal, thinks about the best way to get it done, reaches out to you when it needs your input and recruits the best models and tools to get the job done.
New tools to make your Opal agent more capable
- Memory: Whether it's a user’s name, your style preferences or a running shopping list, your Opals can now remember information across sessions. This makes your Opals grow smarter and feel more personal the more you use them. In this Video Hooks Brainstormer Opal, the agent step stores the user's brand identity and preferences to its memory, allowing you to generate tailored video ideas instantly without repeating your preferences.
- Dynamic routing: Take full command of your workflow by defining multiple paths an agent can follow based on custom logic. Simply describe your criteria, and the agent will intelligently transition to the correct step once those conditions are met. In the Executive Briefing Opal the agent step tailors your briefing based on whether you’re meeting with an existing or new client. It searches the web for new client backgrounds or reviews internal meeting notes to provide relevant context.
- Interactive chat: Sometimes an AI agent needs to ask a follow-up question. The agent step can now initiate a chat with the user to gather missing information, or offer choices before moving to the next stage of the plan. Let's use the Room Styler Opal as an example. If a user didn’t give enough detail at first, the Opal will keep asking questions or showing examples.
More ways to build in Opal
We believe this approach gives you the best of both worlds: the power of an AI agent working towards your goal and the control of a step-by-step workflow you can customize and refine at any time.
We’ve kept Opal simple while making it significantly more capable. New users will find that Opals "just work" because the agent in the generate step is smart enough to self-correct, remember and optimize. For our power users and builders, the standard fixed steps are available whenever you need high-precision prototyping or rigid logic. By bridging the gap between automation and control, we’re expanding the horizons of what you can build. We can’t wait to see your new agent-powered Opals in action!